Research Of Prediction Model In Function Networks And Its Application In Wastewatwe Treatment

Functional networks is a modeling method put forward by the Spanish scholar Enrique Castillo in1998. It is an extension and promotion of the neural network. It is has been widely applied in many fields, and have shown good performance.The thesis explores and researches on functional networks, and carries on the organic integration with neurons split-adjustment mechanism and AIC criterion. It aims to design an efficient nonlinear prediction model for nonlinear prediction systems and finds its application in predicting Pb、Zn、Cd ion concentration and flux at the entrance of wastewater treatment plants.The thesis studied the problem that the current functional structure and basic function are judged from empirical knowledge by human experts and the structure of functional networks and functional parameters are often difficult to determine by manual work. The thesis provided a new model and method for traditional model of functional networks by establishing a standard base function pool which uses neurons split-adjustment mechanism to select the basic function of functional networks and thus determine the structure of functional networks, while taking advantage of the AIC criterion to evaluate, optimize the structure of functional networks, and select the optimal one. Finally, the prediction model based on functional networks is applied to the prediction of ion concentration and flux at the entrance of wastewater treatment plants, and it is compared with the forecast methods of recursive least square and BP neural network by emulate as well. By comparison, we could see that the prediction model based on functional networks to predict the sewage treatment plant entrance ion concentration and flow the relative error of is smaller, higher prediction accuracy, which suggests that the method Can better be used to guide the process of sewage treatment.